Non-linear relationships in covariates

[c777s]

Forgive me as I'm new to CausalImpact and the BSTS modeling on which it's based.

To improve the accuracy of the results, should I split a covariate into multiple tiered covariates to account for non-linear relationships between the covariate and the original time series?

Or does this package already model non-linear relationships in the covariates natively?